Time of occurrence of cardiac death due to arrhythmia, heart failure, or acute myocardial infarction was recorded in 86 elderly subjects, belonging to a group in whom circadian and circannual rhythms in blood pressure and urinary catecholamine excretion had been studied previously. All patients were retired, with no work responsibilities, and lived--closely-supervised in a home for the aged--on a routine that provided little differences between weekdays and weekends. Cardiac mortality showed a circadian variation, with a peak in the early morning hours, coinciding with the circadian peak in systolic and diastolic blood pressures. A weekly (circaseptan) variation in cardiac mortality was found, with the greatest number of patients dying on Mondays and the least on Thursdays. There were seasonal differences in cardiac mortality, with a peak in July and a broader peak during the cold season (December to February). The former coincides with the circannual peak in diastolic blood pressure, but is unrelated to the seasonal variation in norepinephrine excretion. Circadian, circaseptan, and circannual variations in cardiac mortality appear to be the expression of time-dependent, transient risk states for catastrophic cardiac events, which may lend themselves to preventive treatment.